Applying Dynamic Co-occurrence in Story Link Detection

Story link detection is part of a broader initiative called TopicDetection and Tracking, which is defined to be the task ofdetermining whether two stories, such as news articles or radiobroadcasts, are about the same event, or linked. In order to minemore information from the contents of the stories being compared andachieve a more high-powered system, motivated by the idea of theword co-occurrence analysis, we propose our dynamic co-occurrence,which is defined to be a pair of words that satisfy certain relationrestriction. In this paper, relation restriction refers to a set offeatures. This paper evaluates three features: capital, location anddistance. We use dynamic co-occurrence in the similarity computationwhen we apply it in the story link detection system. Experimentalresults show that the story link detection systems based on thedynamic co-occurrence perform very well, which testify the greatcapabilities of the dynamic co-occurrence. At the same time, we alsofind that relation restriction is critical to the performance ofdynamic co-occurrence